t5-small-finetuned-v2-hausa-to-chinese

This model is a fine-tuned version of google-t5/t5-small on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.1509
• Bleu: 30.0183
• Gen Len: 6.4896

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0006
• train_batch_size: 32
• eval_batch_size: 64
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 3000
• num_epochs: 15
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu	Gen Len
1.643	1.0	1103	1.1585	24.9091	6.7771
1.1913	2.0	2206	1.0817	24.5257	6.7541
1.0945	3.0	3309	1.0737	27.3158	6.4568
1.0113	4.0	4412	1.0400	27.6138	6.6673
0.9415	5.0	5515	1.0556	26.3585	6.335
0.8809	6.0	6618	1.0479	25.5111	6.4373
0.8281	7.0	7721	1.0496	26.9639	6.2402
0.7805	8.0	8824	1.0687	28.3541	6.4397
0.7351	9.0	9927	1.0859	28.7719	6.4876
0.6941	10.0	11030	1.1064	27.9477	6.2022
0.6621	11.0	12133	1.1114	29.7176	6.4492
0.6361	12.0	13236	1.1379	29.5086	6.4459
0.6165	13.0	14339	1.1407	29.7825	6.5262
0.6039	14.0	15442	1.1498	30.0064	6.4859
0.6002	15.0	16545	1.1509	30.0183	6.4896

Framework versions

• Transformers 4.45.2
• Pytorch 2.4.1+cu121
• Datasets 3.0.1
• Tokenizers 0.20.1